Deep learning methods with Bayesian-based uncertainty quantification for the emulation of CPU-expensive numerical simulators
In the context of uncertainty propagation in numerical simulations, substitute mathematical models, called metamodels or emulators are used to replace a physico-numerical model by a statistical (or machine) learning model. This metamodel is trained on a set of available simulations of the model and mainly relies on machine learning (ML) algorithms. Among the usual ML methods, Gaussian process (GP) metamodels have attracted much interest since they propose both a prediction and an uncertainty for the output, which is very appealing in a context of safety studies or risk assessments. However, these GP metamodels have limitations, especially in the case of very irregular models. The objective of the post-doctorate will be to study the applicability and potential of Bayesian-based deep learning approaches to overcome these limitations. The work will be focused on Bayesian neural networks and deep GP and will consist in studying their tractability on medium size samples, evaluate their benefit compared to shallow GP, and assess the reliability of the uncertainty associated with their predictions.
Innovative strategies for minor actinides using molten salt reactors
Within the framework of the ISAC (Innovative System for Actinides Conversion) project of the France Relance initiative, preliminary concepts of molten salt reactor capable of incinerating minor actinides have to be proposed in connection with prospective évolutions of the French nuclear fleet (stabilisation or reduction of the plutonium and americium inventory, minimization of the deep storage footprint, …) and contraints linked to the nuclear fuel cycle (plutonium and minor actinides inventories). The specificities of molten salt reactors will be exploited to design innovative transmutation strategies.
The postdoctoral fellow will be based in the reactor and fuel cycle physics unit of the IRESNE R&D institute at CEA Cadarache. He/she will develop expertise in neutronics, fuel physics, and in the design of Generation-IV reactors of the molten salt type.
Design of innovative nuclear systems cooled by heat pipes
The combined goals of CO2 emission reduction and energy self-sufficiency, in the current geopolitical context, open up new perspectives for nuclear applications (cogenerations, hydrogen production, etc.). In particular, the MNR concepts (Micro Nuclear Reactors), with a thermal power of 2 to 50 MW, bear the promise of flexibility, while providing much reliability and accrued safety.
Among the MNR technologies, the particular concept in which the core is cooled by heat pipes strongly improves the inherent safety of the design, in normal and in accidental conditions as well.
In order to demonstrate the feasibility of such an MNR technology, a predesign of a single high temperature heat pipe should be performed for different selected technologies. Then, the overall heat pipe cooling system should be evaluated. Finally, after having modelled the core cooling system, an integration study including a predesign of the core itself should be done with the two subsystems coupled.
New experiments in the PULSTAR research reactor for the validation of high-fidelity multi-physics LWR simulations
The PULSTAR experimental reactor is located at North Carolina State University (NCSU) in the United States. It is a 1 MW pool-type light water reactor fueled with PWR fuel.
The current approach to validate the multi-physics coupling at CEA is exclusively based on experimental data obtained from the operation of the PWR power plants, which is now considered insufficient to validate the coupling of advanced multi-physics models.
In this context, a post-doctoral work is proposed at CEA Cadarache, with the objective of contributing to the specifications of a new experimental program in the PULSTAR reactor for the validation of coupled multi-physics models, as part of a collaboration with the American DoE; the postdoctoral fellow will be jointly supervised by NCSU. The experimental program will study the coupling between multi-physics parameters in steady state conditions, as well as the effects of neutronics/thermal-hydraulics/fuel physics feedback during transients, at the scale of the fuel rod and the water subchannel. Transients of variable dynamics induced by reactivity injection ramps will be considered, at appropriate reactor power levels. An adequate instrumentation to access the local temperatures of the pellet and the cladding should be proposed. The measurements thus collected, under perfectly controlled experimental conditions, will constitute benchmark data for the validation of multi-physics simulations.
The post-doctoral work plan includes the following:
- develop a digital twin of the experimental reactor core, at the fuel rod scale, using CEA simulation tools;
- propose and study configurations of interest by simulation, to contribute to the definition of the experiment for stationary and transient states;
- contribute to the definition of the possible instrumentation for the experiment, according to the target uncertainties.
Depletion calculation of nuclear reactor fuel using Monte Carlo method: moving towards a reference solution
Modern computers offer the possibility to use Monte-Carlo codes to get reference solutions to neutron transport problems. Nevertheless, such reference solutions are only accessible in stationary conditions for practical problems.
The proposed research work aims at exploring and testing methods to obtain a reference Monte-Carlo solution for fuel cycle quantities in depletion problems using present-day computing resources. Such a reference, obtained at a reasonable computational cost, would provide a better control over calculation biases and uncertainties in deterministic solutions typically used in the industry.
Studies will be performed using the Monte Carlo code TRIPOLI-4® coupled with the MENDEL deterministic depletion module. The post-doctoral fellow will perform extensive work on neutron leakage consideration in order to ensure criticality of the model, neutron flux and reaction rates normalization, control of the energy deposition in the different model regions, fine descriptions of the irradiation history, cross section stochastic temperature interpolation, as well as the impact of considering only a limited number of isotopes. Comparisons will be made with the results published by other groups using different approaches and Monte-Carlo codes .
The post-doctoral fellow will be positioned in a team of researchers/engineers in nuclear reactor physics. He/she will improve and deepen his/her knowledge of applied Monte-Carlo simulations as well as the code validation process.
Jet Fuel production plant powered by a nuclear reactor and coupled with a CO2 direct air capture system
As part of the many innovative research projects aiming at achieving carbon neutrality by 2050, one prospective concept is to use a nuclear power plant to produce synthetic jet fuels by sourcing carbon dioxide from the air.
This postdoctoral proposal aims at predesigning an integrated energy conversion system coupling a nuclear energy source with electrochemical and thermochemical processes dedicated to the production of kerosene from water and atmospheric CO2 (Direct Air Capture system). Important considerations for such a system are its energy efficiency, its capacity to meet projected industrial needs and its competitiveness in a future decarbonized energy market.
The postdoctoral fellow will first perform numerical simulations in order to sketch an optimized process flow configuration coupling the DAC system with the nuclear power plant. He/She will also define the operational conditions of the system. In a second step, in association with another postdoctoral fellow working on the chemical transformation processes, he/she will perform a global integration of the various processes involved, from the atmospheric dioxide carbon to the final jet fuel, taking into account the required flows of heat and electricity. The design drivers will be the optimization of the overall plant efficiency and the operational conditions. Finally, he/she will propose a preliminary balance of plant accounting for regulatory constraints in order to evaluate the main design factors such as the required CO2 capture surface area for example.
The postdoctoral fellow will be based in a research unit specialized in innovative nuclear system studies. He/She will develop a technical understanding of prospective nuclear and decarbonization technologies combined.
Analysis of the SEFOR experiments for the multi-physics validation of fast reactor simulation tools :
In the Verification Validation and Uncertainty Quantification process of modern simulation tools, the validation phase relies mainly on the comparison between calculation and experimental results for the major quantities of interest. For neutronics, the experiment database focuses on measurements coming from zero power reactors for which the reference states does not require complex multi-physics modeling: isothermal state (very low power such as few hundreds of watts) and fresh fuel (un-irradiated).
However, the VVUQ of power reactor needs to go beyond zero power experiments and thus arises the necessity to apply a multi-physics VVUQ approach. This new frame requires the integration of phenomena from other disciplines outside of pure neutronics: temperature and density dependence of the main quantities of interest (keff, power distribution, and feedback coefficients), temperature field inside the pins as function of core power and irradiation.
Regarding Doppler Effect, the set of experiments held at the SEFOR facility in the 70’s is of major interest for the VVUQ process. This sodium cooled fast reactor fed with mixed oxide fuel was built in support of the US R&D program for indigenous code validation at the time.
Based on the available data, the proposed work focuses on core characterization using a fully neutronic/thermo-mechanic/thermalhydraulic process for both nominal and transient states based in high-fidelity modeling. In order to quantify the benefit of such approach, a step-by-step comparison will be done with the same results obtained by the traditional “chained approach” which assumes a weak dependence between the three mentioned disciplines.
The work will be performed using the last generation of simulation tools available at CEA.